Retrieval of average CO2 fluxes by combining in situ CO2 measurements and backscatter lidar information

Retrieval of average CO2 fluxes by combining in situ CO2

measurements and backscatter lidar information

Fabien Gibert,1 Martina Schmidt,2 Juan Cuesta,1 Philippe Ciais,2 Michel Ramonet,2

Irene Xueref,2 Eric Larmanou,3 and Pierre Henri Flamant1

Received 27 October 2006; accepted 1 February 2007; published 16 May 2007.

[1] The present paper deals with a boundary layer budgeting method which makes use ofobservations from various in situ and remote sensing instruments to infer regional averagenet ecosystem exchange (NEE) of CO2. Measurements of CO2 within and above theatmospheric boundary layer (ABL) by in situ sensors, in conjunction with a preciseknowledge of the change in ABL height by lidar and radiosoundings, enable to inferdiurnal and seasonal NEE variations. Near-ground in situ CO measurements are used todiscriminate natural and anthropogenic contributions of CO2 diurnal variations in theABL. The method yields mean NEE that amounts to 5 mmol m�2 s�1 during the night and�20 mmol m�2 s�1 in the middle of the day between May and July. A good agreement isfound with the expected NEE accounting for a mixed wheat field and forest areaduring winter season, representative of the mesoscale ecosystems in the Paris areaaccording to the trajectory of an air column crossing the landscape. Daytime NEE is seento follow the vegetation growth and the change in the ratio diffuse/direct radiation. TheCO2 vertical mixing flux during the rise of the atmospheric boundary layer is alsoestimated and seems to be the main cause of the large decrease of CO2 mixing ratio in themorning. The outcomes on CO2 flux estimate are compared to eddy-covariancemeasurements on a barley field. The importance of various sources of error anduncertainty on the retrieval is discussed. These errors are estimated to be less than 15%;the main error resulted from anthropogenic emissions.

Citation: Gibert, F., M. Schmidt, J. Cuesta, P. Ciais, M. Ramonet, I. Xueref, E. Larmanou, and P. H. Flamant (2007), Retrieval of

average CO2 fluxes by combining in situ CO2 measurements and backscatter lidar information, J. Geophys. Res., 112, D10301,

doi:10.1029/2006JD008190.

1. Introduction

[3] Ecosystem carbon exchanges on spatial scales ofapproximately 1 km2 can be well documented using theeddy-covariance (EC) technique [e.g., Wofsy et al., 1993;Baldocchi et al., 1996; Valentini et al., 2000]. However, afew data exist on the carbon budget at the regional scale,ranging from a few tenths to a few hundreds of squarekilometers. At these scales, the carbon fluxes can beestimated either by upscaling pointwise flux measurementsusing for instance airborne flux transects, biophysicalmodels, and remote sensing information [Turner et al.,2004, Miglietta et al., 2006] or by inverting atmosphericconcentration measurements using a fine-scale mesoscaletransport model (M. Uliasz, et al., Uncovering the lake

signature in CO2 observations at the WELF tall tower: Amodelling approach, submitted to Agricultural and ForestMeteorology, 2006). This latter method however is stillunder development and requires very dense atmosphericobservation data sets (A. J. Dolman, et al., CERES: TheCarboEurope Regional Experiment Strategy in Les Landes,South West France, May–June 2005, submitted to Bulletinof the American Meteorological Society, 2006).[4] In the near future, passive remote sensing instruments

will be launched soon (OCO, GOSAT) [Crisp et al., 2004;Inoue, 2005] to measure column CO2 concentration and toconstrain regional flux inversions. In this context, a differ-ential absorption lidar (DIAL) has been developed at theInstitut Pierre Simon Laplace (IPSL)-Laboratoire de Meteo-rologie Dynamique (LMD) to conduct simultaneous meas-urements on CO2 density and atmospheric boundary layer(ABL) structure [Gibert et al., 2006]. The two sets ofmeasurements provide information on CO2 reservoirs andmixing processes during the diurnal cycle.[5] The goal of this paper is to infer surface CO2 fluxes at

larger scale than EC measurements by estimating the pro-cesses that control the CO2 concentration variations in theatmospheric boundary layer (ABL). Several analyses of theCO2 mass balance of the ABL have already been reported to

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, D10301, doi:10.1029/2006JD008190, 2007ClickHere

for

FullArticle

1Institut Pierre Simon Laplace, Laboratoire deMeteorologie Dynamique,Ecole Polytechnique, Palaiseau, Cedex, France.

2Institut Pierre Simon Laplace, Laboratoire des Sciences du Climat et del’Environnement, UMR CEA/CNRS 1572, Gif-sur-Yvette, Cedex, France.

3INRA Unite Mixte de Recherche INRA/INAPG ‘‘Environnement etGrandes Cultures’’, Thiverval-Grignon, France.

Copyright 2007 by the American Geophysical Union.0148-0227/07/2006JD008190$09.00

D10301 1 of 16

estimate regional-scale fluxes, usually with aircraft verticalprofiles [Levy et al., 1999; Lloyd et al., 2001]. These studiesstressed that accurate estimates of the ABL height, structure,and evolution and information on horizontal advection areessential to determine the surface fluxes. In most cases, theABL budgets were measured only during the day, and thelack of nighttime information is problematic to estimaterespiration sources [Lloyd et al., 2001].[6] Elastic scatter lidar has already showed its abilities to

monitor the diurnal cycle of the ABL [Menut et al., 1999]. Inthis paper, a new ABL mass balance method is developed toretrieve land biotic regional fluxes of CO2 in the southwest ofParis during one growing season. The study covers Marchthrough September 2004, when the land biotic fluxes domi-nate anthropogenic emissions in controlling CO2 variationsin the ABL. The method uses ground-level in situ continuousCO2 observations, airborne CO2 vertical profiles reachinginto the free atmosphere, radiosoundings, and lidar measure-ments (Figure 1). Solar flux pyranometer measurementsare also used. We seek to provide separate estimates ofthe nighttime and daytime regional mean net ecosystemexchange (NEE) and of the flux of CO2 due to entrainment atthe top of the ABL (M). The regional NEE inferred fromABLobservations is compared with EC flux measurementsconducted over a barley field.

2. Processes That Drive CO2 Variations in theBoundary Layer

2.1. CO2 Balance in the ABL

[7] The mean CO2 mixing ratio in the ABL changes inresponse to land biotic fluxes, dynamical processes, and

anthropogenic emissions. A mass balance approach shows thatthe rate of change in the CO2 mean mixing ratio in the ABL,hCi, is driven by the sum of the fluxes across the boundaries ofan air column extending from the ground to the top of the layerconsidered. Accordingly, the evolution of hCi is described bythe following equation [Lloyd et al., 2001;Gerbig et al., 2003]:

rhd Ch idt

¼ NEE þM þ A ð1Þ

where r is the mean molar air density (mole per cubicmeter), h is the ABL height (meter), C(z, t) is the mixingratio profile of CO2 in the ABL, NEE is the mean regionalnet ecosystem exchange (mole per square meter persecond), and M is the molar flux per surface unit due tovertical mixing with the upper layer, either the residual layer(RL) or the free atmosphere (FA). NEE is representative ofthe heterogeneously distributed fluxes that are transportedby mesoscale flow. We consider that the ABL spatiallyintegrates surface fluxes over areas with length scales muchlarger than h [Gloor et al., 1999]. A is the flux of CO2 fromadvected anthropogenic emissions.[8] The value of hCi and M are given by:

CðtÞh i ¼ 1

h

Zh

0

Cðz; tÞ dz ð2Þ

M ¼ rdh

dtCþ � Ch ið Þ ð3Þ

where C+ is the CO2 mixing ratio in the layer above theABL. The term dh/dt accounts for the vertical velocityinduced by large-scale atmospheric motion as subsidenceand by the ABL growth and decay controlled by sensibleheat flux. In other studies [Lloyd et al., 2001; Gerbig et al.,2003], subsidence has been added explicitly in equation (3).Here we included it in equation (3) because of the nature ofour h observations (section 4.2).[9] Daytime NEEday is assumed to decrease linearly with

the total short-wave solar flux (see section 6):

NEEday ¼ �aLþ b ð4Þ

where L is the total (direct + diffuse) short-wave down-welling solar flux (watt per square meter) measured by apyranometer, and b is the NEE value at the sunrise. We calla the light conversion factor which reflects biologicalprocesses (micromole per square meter per second/watt persquare meter).

2.2. Characteristic Diurnal Regimes

[10] Figure 2 shows typical summertime diurnal varia-tions of CO2 measured near the ground using a gaschromatograph and of the ABL thickness h measured fromelastic backscatter lidar measurements, together with thesolar flux from pyranometer measurements (see section 3).Three distinct regimes can be seen.[11] 1. During the night, the CO2 mixing ratio usually

increases linearly from 2000 to 0500 UT. The nocturnalboundary layer (NBL) behaves as a reservoir filled with thenocturnal plant and soil respiration emissions. Assuming

Figure 1. Measurement area with CO2 and CO in situground-based measurements at LSCE laboratory. Lidar andpyranometer (L) measurements at LMD laboratory and radio-soundings (RS) at Trappes andLMDfacilities. Eddy-covariance(EC) system to measure CO2 flux above crops at INRA facility.

D10301 GIBERT ET AL.: CO2 FLUX

2 of 16

D10301

anthropogenic emissions to cause only a weak disturbancefrom the respiration-driven linear increase of CO2 (seesection 4.1) and knowing the NBL height (hNBL), thenighttime NEE flux is given by:

NEEnight ¼ rhNBLd Ch idt

ð5Þ

[12] 2. Between the sunrise and the occurrence of themixed layer (ML) at 0800 UT, both respiration and photo-synthesis influence the rate of change of near-ground CO2. Atthis time of the day, a strong temperature inversion oftenremains from the former night, which is not disturbed by solarheating (L < 500 W m�2). Therefore, assuming that h = hNBLduring this morning regime, the daytimeNEEday is also givenby equation (5).[13] 3. Between approximately 0800 and 1400UT, the near-

ground CO2 mixing ratio is decreasing due to photosynthesisand vertical mixing with the residual layer (RL) and eventuallythe free atmosphere (FA) above (see arrows in Figure 2c). TheCO2mixing ratio iswellmixed inside the convective boundarylayer. At the end of the afternoon, the near-ground CO2

mixing ratio becomes nearly equal to the value in the freetroposphere. All fluxes in equation (1) are active. However, asthe ML is more than 1000 meters thick, the surface fluxeschange the CO2 mixing ratio only slowly with time.

2.3. Contamination by Anthropogenic Emissions

[14] The measurement sites are located 20 km at southwestof Paris in a rural area (Figure 1). Figure 3 displays theaverage near-ground CO2 and CO diurnal cycles observedbetween April and September, together with the mean roadtraffic index in Paris urban area (SYTADIN data are availableat http://www.sytadin.equipement.gouv.fr). Car trafficaccounts for �60% of the Paris CO2 emissions during thatperiod (CITEPA and AIRPARIF data are available at http://citepa.org and http://www.airparif.asso.fr, respectively).Therefore CO can be used as a car traffic marker to evaluatethe anthropogenic contamination of the CO2 diurnal varia-tions. Depending on the wind direction, an increase in theCO2 mixing ratio is expected during the rush hours between0600 and 0800 UT in the morning (or 0800 and 1000 UTlocal summertime) and between 1600 and 1800 UT in theevening. The effect of emissions on ABL CO2 mixing ratios

Figure 2. 30 July 2004: (a) Pyranometer solar flux measurements. (b) CO2 ground-based in situmeasurements at 10 m height (IPSL/LSCE). Biological and dynamical processes that govern CO2 diurnalcycle have been added: NEE, net ecosystem exchange; M, mixing. (c) Lidar 532-nm boundary layeraerosol backscatter signal. The boundary layer diurnal cycle is also represented: RL, residual layer; NBL,nocturnal boundary layer; ML, mixed layer; FA, free atmosphere [Stull, 1988]. The double arrowsindicate the vertical mixing with, first, the RL and, second, the FA. The simple arrows are for the dailyTrappes radiosoundings (RS).

D10301 GIBERT ET AL.: CO2 FLUX

3 of 16

D10301

is stronger in the morning when the ABL is shallow. Braud etal. [2004] have shown that the anthropogenic emissions ofParis are characterized by a mean ratio CO2/CO of 0.09(11 ppb of CO per parts per million of fossil CO2). This meanemission flux ratio should only be used as an approximationfor the actual ratio of fossil CO2 to CO in the atmospherewhich depends on the distance between the measurement siteand the diverse combustion sources as well as on localatmospheric transport. When using the fossil CO2 to COemission ratio as qualitative indication of the amount of fossilCO2 in the ABL, we found that the diurnal cycle of CO2 isoften contaminated by urban emissions between 0700 and1200 UTwhile it is rather insensitive to them in the middle ofthe afternoon. This ‘‘fixed zone’’ corresponds to approxi-mately 2 hours or �20 CO2 measurements (Figure 3).

3. Experimental Set up

3.1. Study Area and Measurement Sites

[15] Twenty-one days has been analyzed to infer surfacefluxes. Figure 1 shows the Paris urban area and the measure-ment sites, placed 20 km southwest of the most denselypopulated area. Quasi-continuous CO2 and CO near-ground

in situ measurements are made at the IPSL-LSCE site. Solarflux and lidar measurements are made at the IPSL-LMD site,distant of LSCE from 5 km. A pyranometer at IPSL-LMDprovides total (L), direct, and diffuse solar flux observations.Meteorological sensors are in place at both sites. Verticalprofiles of humidity, temperature, and horizontal wind speedare obtained from daily radiosoundings at Trappes (Figure 1)at 1100 and 2300 UT. The Trappes meteorological facility islocated �10 km to the west of the LSCE and LMD labora-tories. At such a distance in the Paris plains, terrain-inducedinhomogeneities are not expected to have a major impact onthe mean ABL structure and parameters. Additional radio-soundings at IPSL-LMD made during May and June 2004confirmed this hypothesis (see section 7).

3.2. CO2 and CO In Situ Measurements

[16] An automated gas chromatographic system (HP-6890)was routinely operated at IPSL-LSCE for CO, CO2, CH4,N2O, and SF6 semicontinuous measurements of ambient air[Worthy et al., 1998]. The precision on CO2 is ±0.5 ppmwith ameasurement time step of 5 min [Pepin et al., 2002] as shownin Figure 4. Additionally, airborne vertical CO2 profiles havebeen conducted once every 15 days using flask sampling and acontinuous LICORNDIRgas analyzer onboard a light aircraft.These routine flights are made over the Orleans forest (80 kmsouth of the study area). They reach up to 3000 meters,piercing the top of the ABL and sampling the lowermost freeatmosphere (Figure 4).

3.3. Ecosystem CO2 Fluxes

[17] Eddy-covariance NEE fluxes were measured 10 kmaway from the IPSL-LMD and IPSL-LSCE sites above abarley canopy (INRA data are available at http://www.inra.fr)since May 2004. The barley winter crops were growing atthat time and gradually dried up afterward until harvest tookplace on 2 July. After harvest, the measurements went onabove a thatch field up until September.

3.4. Lidar Measurements of ABL Structure

[18] Elastic lidars pointing at zenith are used to monitor theABL structure and diurnal changes. Aerosol particles arereliable tracers of the ABL height and fluctuation. The rangeresolution is 15 m, and a profile is acquired every 10 s(Table 1). The ML and RL layers are respectively identifiedby their large and small aerosol concentrations (Figure 2). Thelidar configuration precludes measurements below 200 m dueto a lack of overlap between the laser transmitter and thetelescope receiver. Therefore we use vertical profiles of poten-tial temperature from Trappes radiosoundings to determine theNBL height. During the day, both lidar and radiosoundingobservations are used to retrieve the height of the ML.

4. Flux Retrieval Method

4.1. Retrieval of CO2 Fluxes

[19] Figure 5 shows how the different data sets were usedto retrieve NEE and the ABL entrainment flux M. The valueof M is determined by combining measurements of dh/dt(see section 4.2) with near-ground CO2 concentrations andairborne flask CO2 samples in the free atmosphere. The CO2

mixing ratio in the RL is assumed to be equal to that of theML at the end of the afternoon of the previous day (obtainedby near-ground in situ data). The free tropospheric CO2

Figure 3. Mean ground-based CO2 (solid line) and CO(dashed line) in situ measurements between April andSeptember 2004. The time resolution is 5 min. The greyareas correspond to the standard deviation. The meananthropogenic CO2 is calculated using CO measurements(assuming that 10 ppb of CO corresponds to 1 ppm ofanthropogenic CO2, i.e., h = 0.1) and then is deducted fromthe whole CO2 diurnal cycle to retrieve the CO2 naturalvariations (dashed and dotted line). The ‘‘fixed zone’’considered to constrain regional flux retrieval is indicatedas a rectangle and corresponds to �20 points or 2 hours ofCO2 measurements in the mid-afternoon. Linear fit y = at + bis used to retrieve the respiration flux during the night. Themean road traffic index in Paris urban area during this periodof time is also displayed (SYTADIN data are available athttp://www.sytadin.equipement.gouv.fr).

D10301 GIBERT ET AL.: CO2 FLUX

4 of 16

D10301

value is estimated each day using a time interpolation of insitu airborne measurements (Figure 4).[20] TheNBLheights are determined from radiosoundings at

2300 (UT). Combined with hourly near-ground CO2 data, theNBL height is used to determine NEEnight using equations (2)and (5). Figure 3 shows that, on average, the linear increase ofCO2 during the night implied by NEEnight is only weaklydisturbed by anthropogenic sources. This is due both to theprevalence of well-mixed boundary layers in the late afternoonand to the very small CO2 emissions at night. Fitting a lineartrend to the CO2 mixing ratio (y = at + b) thus is possible andallows to infer NEEnight. The accounted for remaining anthro-pogenic emissions create statistical errors in the retrieval of aand b and add a systematic bias on the CO2 diurnal cycle (db).4.1.1. Direct Method[21] In rare occasions when the CO2 diurnal cycle is not

disturbed by anthropogenic emissions, we could directlydetermine NEEday using the full time period between thesunrise and the ABL rise using equation (5). The value of awas then used to determine NEEday during the rest of theday using equation (4) (see Figure 5).4.1.2. Iterative Method[22] In the vicinity from Paris urban area, the morning

period is most often disturbed by anthropogenic emissions.Consequently, we cannot use equation (5) to determine a in85% of the cases studied. KnowingM and b (NEEnight at thesunrise), we thus looked for a value of a = (NEEday � b)/Lthat matches best the observed CO2 decreasing trend duringthe ‘‘fixed zone’’, lasting for �2 hours in the late afternoon

(see Figure 3). NEEday is determined after several iterationsadjusting a in order to reach the observed trend for CO2

during the late afternoon. For every iteration, M gets alsoadjusted using the modeled natural CO2 diurnal cycle inequation (3). This iterative method does not require anyhypothesis about fossil CO2 to CO ratios which may be veryvariable (section 6.4).[23] Without anthropogenic emissions, both the direct and

the iterative methods to retrieveNEEday are applicable and canbe compared. This comparison is performed in section 6.2.Moreover, in the iterative method, additional information onthe ratio between fossil CO2 and CO in the air, h, can beobtained from CO in situ measurements (section 6.4).

4.2. Retrieval of ABL Height and hhhhCiiii Using Lidar

Backscatter Signal and Radiosoundings

[24] The inflexion point method (IPM) [Menut et al., 1999]is used to determine the ABL height from radiosoundings(night) and lidar backscatter signal (day). Figure 6 displaysthe lidar backscatter signal and the radiosounding potential

Figure 4. (a) CO2 in situ ground-based measurements at LSCE facility (gray solid line) and freetropospheric airborne measurements 100 km south of Paris for the year 2004 [squares for eachmeasurement and interpolation (solid line)]. (b) CO in situ ground-based measurements at LSCE facilityfor the year 2004 (gray solid line). The black solid lines correspond to the 21 cases studied in this paper.

Table 1. Elastic Scattering Lidar Characteristics at LMD Facilitya

Laserl,nm

E,mJ

PRF,Hz

Telescope D,cm/FOV, mrad

DR,m

Averaging,s

Nd:YAG 532 30 20 20/3 15 101064 15

al is the emission wavelength, E is the average output laser energy, D isthe telescope diameter, FOV is the field of view, and DR is the verticalresolution.

D10301 GIBERT ET AL.: CO2 FLUX

5 of 16

D10301

temperature profiles from Trappes and LMD on 25–26 May.The 0157 UT potential temperature profile shows a largetemperature inversion close to the ground, which caps theshallow NBL. This NBL height is determined from themaximum of the second-order derivative of q. The samemethod is used to determine the top of the RL.[25] During the night, knowledge of the q vertical profile

is usually sufficient to characterize vertical stratification andscalar gradients in the boundary layer [Stull, 1988]. Studiesof CO2 and H2O vertical profiles show that mixing ratiosand potential temperature profiles usually have a similarshape due to nocturnal static stability [Bakwin et al., 2003;Lloyd et al., 2001; de Arellano et al., 2004; Dolman et al.,2005]. In order to determine the mean CO2 mixing ratio inthe NBL, we assume a similarity between the verticalgradients of C and q, the gradient of q being assumed tobe constant during the whole night. This yields:

C z; tð Þ¼CRLþ C 0; tð Þ � CRL½ qð0Þ�qðzÞqð0Þ�qðhNBLÞ

� �ð6Þ

where CRL is the CO2 density in the residual layer (hNBL < z <hRL), and C(0, t) is the observed near-ground CO2 mixingratio. The mean CO2 mixing ratio in the NBL hCi is thenobtained by solving equations (2) and (6). Errors associatedto the similarity hypothesis will be analyzed in section 7.2.[26] During the day, when the ABL is mixed by convec-

tion (Figure 2), CO2 can be assumed to be constantvertically and equal to the near-ground in situ value. ThushCi = C(0, t). The convective ABL height is determined bythe inflexion point method minimizing the second deriva-tive of backscatter lidar signal with respect to altitude@2(Pz2)/@z2. The mixed-layer height hML is then definedas the middle of the transition zone between the mixed layer

and the free troposphere (see Figure 6). The ML heightsdetermined either from the lidar signals or from radio-sounding profiles are comparable within 5%.

5. Results

5.1. Diurnal Trends Controlling CO2 in the ABL

[27] The inferred NEE and mixing CO2 fluxes are shownin Figure 6c for the period 25–26 May. The flux M is thelargest for a maximum difference in CO2 mixing ratiobetween ML and RL and when the ABL rate of growth ismaximal. During the growth of the ABL, thermal over-shoots entrain air from, first, the residual layer and, second,from the free atmosphere. This entails a significant dilutionof the ABL concentrations in trace gases. Thus mixingcombined with daytime NEE uptake causes a large drop ofCO2 in the morning. Remark that CO2 is decreased muchmore effectively by vertical mixing than by NEE during themorning growth of the ABL, as also observed by Yi et al.[2000], Davis et al. [2003], and de Arellano et al. [2004]. Inthe afternoon, NEE persistently removes CO2 from the ABLbut results only into a small trend of the mixing ratio value(�1.1 ppm/hr) because the height of the ABL is maximum.The case of 26 May is not ideal because a rainfall occurredat 1600 UT with dense clouds reducing the NEE uptake ofCO2 (Figures 6a and 6c). During the afternoon well-mixedABL regime, it is worth noting that M is slightly positive(see, for example, Figure 6c for 25 May and Figure 7 for27 April), meaning that CO2 enters from the free tropo-sphere into the boundary layer.[28] Once the NEE parameters (NEEnight and a) are deter-

mined, we can model the full diurnal variation of CO2

(caused by biotic fluxes only) in the ABL. For the ‘‘cleandays’’, without anthropogenic emissions, the modeled dropof land biotic CO2 in the morning is in good agreement with

Figure 5. Block diagram and method to retrieve CO2 surface fluxes due to biological (NEE) anddynamical (M) processes. Circles are for the measurements while squared frameworks are for the results.The dotted arrows are used whenever the CO2 diurnal cycle is not disturbed by anthropogenic emissions.The double arrows are for successive iterations and retroactions. hNBL and hML are the nocturnal and themixed-layer height, respectively, q is the potential temperature, a is the light conversion factor, a and bare the linear fit coefficients described in Figure 3, and h is the ratio between fossil CO2 and CO.

D10301 GIBERT ET AL.: CO2 FLUX

6 of 16

D10301

the observations, which validates our assumptions of homo-geneous mixing of regional fluxes and NEEday being propor-tional to solar radiation (19March and 30 July in Figure 7). Forthe ‘‘polluted days’’, on the other hand, an extra fossil CO2

component is needed in the ABL budget to fit the morningCO2 drop (27 April and 1 September in Figure 7).

5.2. Seasonal Variations

[29] The results are given in Figure 8. The seasonalevolution of the flux M reflects seasonal variations in theABL rate being linked to the stratification of the ABL duringthe night (strength of temperature inversion) and to the

turbulent energy transfer with the air above [Stull, 1988]. Aseasonal maximum inM amplitude can be seen in the middleof the year when the surface sensible heat flux reaches amaximum (see Mmin values in Figure 8a and H values inFigure 8b). In summer, the flux M is a sink for CO2 in theABL during the morning and a source during the afternoon,when the tropospheric CO2 value is higher than the ML one,due to vegetation uptake. More precisely, Figure 8a showsthatM is a net daily source of CO2 for the ABL in May, whenCO2 value remains higher in the free troposphere than in theABL [see also de Arellano et al., 2004; Yi et al., 2004;Bakwinet al., 2004; Hellinker et al., 2004].

Figure 6. 25 and 26 May 2004: (a) Backscatter lidar signal. Color plot is for Ln(Pz2) in arbitrary unit(red is for higher return signal). (b) NBL and ML height retrievals from the inflexion point method (IPM).The ABL height is displayed using backscatter lidar measurements (fine solid line) and is smoothed toretrieve the mean ABL height (bold red solid line). Potential temperature profiles from radiosoundings,launched at LMD (black solid line) and at Trappes (blue solid line) facilities, are displayed in arbitraryunits. These profiles are used to calculate the mean NBL height. (c) Flux retrievals. NEE is the netecosystem exchange from the respiration and the photosynthesis uptake, and M is the flux due to thevertical mixing. Total CO2 mixing ratio from in situ measurements (black dashed line) and land bioticCO2 (obtained using a linear correction by CO measurements) (green solid line) and modeled CO2

diurnal cycles (blue solid line) are also displayed. The rectangles correspond to the fixed zones used tocompute the modeled natural CO2 diurnal cycle.

D10301 GIBERT ET AL.: CO2 FLUX

7 of 16

D10301

[30] Between April and June, both daytime and nighttimeNEE amplitudes reach their seasonal maximum (Figure 8a)in good agreement with the local growing season and withthe eddy-covariance measurements above forests and cropsin western Europe [Granier et al., 2002; Perrin et al., 2004;Moureaux et al., 2006]. Day-to-day variations in NEE arefurther investigated in sections 6.2 and 6.3.[31] The standard deviation and themean difference between

modeled (Cm) and observed (Ce) daily CO2 mixing ratios aredisplayed in Figure 8c. Even for polluted days associated withhigh levels of CO (27 April, 11 May, and 1–2 September), themodeled diurnal fluctuation of land biotic CO2 is in goodgeneral agreement with the in situ data corrected for anthropo-genic effects (Ce� hd(CO)). The mean difference of the meanshCe� hd(CO)� Cmi equals�0.6 ppm, and the mean standarddeviation between the model and the corrected data s(Ce �hd(CO)� Cm) is 2.1 ppm. For the clean days less contaminatedby anthropogenic sources (19 March, 2 July, 30 July, and26 September), the model produces its best agreement withthe data, and the mean standard deviation is below 1 ppm.

6. Discussion

6.1. Comparison With Eddy-Covariance FluxesMeasurements

[32] In the region study, agriculture (mainly winter wheat)represents 50% of the land use while forests (mostly oaks)cover 30% (CHAMBAGRI and IFN data are available at

http://www.ile-de-france.chambagri.fr and http://www.ifn.fr,respectively), the rest being urban. Crops usually have higherdaytime NEE uptake rates than forests, but their growingseason is shorter. We compared the inferred regional NEEto the Barley eddy-covariance data (see section 3.3) inFigure 8d. The eddy-covariance NEE exhibits the highestphotosynthesis rates in mid-May and lower values after-ward when the plants begin to dry. After harvest on 2 July,the thatch field releases CO2 into the atmosphere.[33] The scales at which ABL budgeting and eddy-

covariance methods measure NEE are so different that adirect comparison is not feasible. However, we can look forsimilarities in seasonal and synoptic NEE changes even ifwe expect different values. The ABL-inferred NEE has aquite different phase than eddy-covariance data. Theregional NEE keeps negative values after the harvest andis therefore influenced by other ecosystems than just wintercereal fields. A zoom on the 25–26 May period (Figure 8d)compares ABL and eddy-covariance NEE. The daytimevalues are nearly the same (except for 26 May maybebecause of cloudiness differences). The nighttime respira-tion values are markedly higher in the NBL budgetingmethod (5 mmol m�2 s�1) compared to the flux towermeasurements (2 mmol m�2 s�1) possibly because cultivatedsoils contain less carbon than forest soils, influencing theNBL trends. In September, a similar negative trend ofnighttime NEE between NBL and eddy-covariance method

Figure 7. NEE and M retrievals and CO2 diurnal cycle for four cases in the year 2004: 19 Mar, 27 Apr,30 Jul, and 1 Sep. In situ CO2 ground-based measurements from LSCE facility (black dashed line) andapproximated CO2 diurnal natural variations using a linear correction from CO measurements (grey solidline). The black solid line is for the model.

D10301 GIBERT ET AL.: CO2 FLUX

8 of 16

D10301

can be seen. This decrease is parallel to the decrease of theheat and the latent fluxes (Figures 8b and 8d).

6.2. NEE Response to Solar Flux

[34] We have assumed in equation (4) that NEEday

increases linearly with incoming short-wave solar flux. Tofurther test this hypothesis, we regressed the eddy-covarianceNEE as a function of the solar flux in May. Even under clear-

sky conditions with high solar flux (for example, up to1000 W m�2 on 25 May), the daytime NEE uptake wasalways observed to increase linearly with L (Figure 9a).Such a linear behavior is observed for crop canopies [e.g.,Anthoni et al., 2004a], but not for forests, where NEE oftensaturates with light [Granier et al., 2002; Anthoni et al.,2004b].

Figure 8. (a) Fluxes retrievals obtained for the 21 cases studied in the year 2004. NEE is the netecosystem exchange [cross (minimum values), filled circles (mean daily values), times cross (maximumvalues)], and M is the flux due to the vertical mixing [empty triangles (maximal and minimal values) andfilled squares (mean daily values)]. (b) Sensible Q (solid line) and latent H (dashed line) heat fluxesmeasured at INRA facility. (c) Difference of the mean daily CO2 mixing ratios from the ground-basedmeasurements [Ce and Ce � hd(CO)] and the modeled diurnal cycles (Cm). The daily standard deviationsbetween Ce and Cm and between Cm and Ce � hd(CO) are also displayed. (d) Inferred NEE from theboundary layer budget and measured using an eddy-covariance system above a barley field at INRAfacility. INRA NEE measurements begin on 15 May. The gap in July is for the harvest period. A zoom isdisplayed for the two cases 25 and 26 May 2004.

D10301 GIBERT ET AL.: CO2 FLUX

9 of 16

D10301

[35] For the ‘‘clean days’’, by using equation (5), we candetermine directly the regional NEE dependency on solarflux for the morning period (Figure 9b). This inference islimited in time between the sunrise and the onset ofconvection. The days of 30 July and 26 September arebest suited for this exercise, being almost exempt ofanthropogenic contamination during the morning and char-acterized by a late onset of convection. During these 2 days,a highly significant linear relationship between NEE and Lwas found (the lower correlation on 30 July being due topartly cloudy conditions as seen in Figure 2).[36] Figure 10 further shows the seasonal variation of a

inferred from the ABL budgeting method. A maximum isseen in May during the peak of the growing season. Theerror bars on a reflect errors on the method and on thedetermination of the afternoon fixed zone. The potentialcontamination of the CO2 trends during the afternoon fixedzone by anthropogenic emissions could also induce asystematic error on a. For clean days, we can comparethe ‘‘direct’’ and ‘‘iterative’’ methods to determine a duringthe morning. The values of a obtained by both methods areshown in Figure 10, showing their good agreement (meandifference of ±11%).[37] Several studies have shown that canopy light-use

efficiency (proportional to our light conversion factor a) ishigher under diffuse radiation conditions [Gu et al., 2002;Min, 2005]. We hence investigated the dependency of dailya upon the diffuse fraction f of the solar flux (Figure 11a)and found a significant positive correlation between a and f(r = 0.76). Variations in the diffuse fraction of light werefound to explain 36% of the variance of a between Marchand July. In addition, daily variations in L are anticorrelatedwith a (r = �0.62) as shown in Figure 11b. Higher net solarfluxes are usually associated to clear-sky conditions (smallvalues of f) and to a decrease of the relative humidity in theABL. Figure 11c shows that a is weakly correlated withrelative humidity changes (r = 0.36), possibly suggestingeffects of plant water stress.

6.3. Nighttime Respiration Process and Response toAir Temperature

[38] Temperature is a commonly studied environmentalcontrol on soil respiration [Lloyd and Taylor, 1994; Kattereret al., 1998; Granier et al., 2002; Perrin et al., 2004;Reichstein et al., 2003]. Soil temperature increasingly lagsair temperature deeper into the soil, but a significantpositive correlation between air temperature (Tair) andrespiration is still observed at flux tower sites, especiallybetween May and September when the discrepanciesbetween air and soil temperatures are reduced [Moureauxet al., 2006]. Figure 12 shows a weak correlation betweenour inferred respiration (NEEnight) and air temperature [r =0.43; p = (0.35 ± 0.17) mmol m�2.s�1/�C in Figure 12a].

Figure 9. NEE as a function of the total radiative solar flux L: (a) Measured at INRA facility above abarley field on 25 May 2004; (b) estimated from equation (5) and LSCE ground-based CO2

measurements in July and September during the period of time (2) (see Figure 2).The linear correlationcoefficient (r) and the absolute value of the slope (a) [mmol m�2 s�1/(W m�2)] are indicated.

Figure 10. Light conversion factor, a (mmol m�2 s�1/(W m�2)], as a function of time for the several casesstudied during the year 2004 using the iterative method(grey circles) and using the direct method (Figure 5, dottedarrows) (black diamonds). A 4-point sliding averaging ona estimates is also displayed (dashed line).

D10301 GIBERT ET AL.: CO2 FLUX

10 of 16

D10301

The linear slope of respiration versus Tair is larger inApril–May than during July–September possibly becauseof higher soil water content in spring [Moureaux et al.,2006].

6.4. Estimates of Regional Fossil CO2 to CO EmissionRatios

[39] We can calculate for each day the hourly fraction ofland biotic CO2 and can deduce by difference from the

observations the fraction of anthropogenic CO2. This makesit possible to estimate daily values of the ratio h of fossil CO2

to CO. This ratio may vary in space and time depending onthe nature and heterogeneity of anthropogenic sources(higher temperature combustions emitting less CO relativeto CO2) and on the footprint of the atmospheric measurementsite with respect to diverse CO and CO2 sources. Fortunately,the semirural IPSL-LSCE site is far enough from immediatepollution sources. The daily values of the ratio h, as a function

Figure 11. Normalized variations of the light conversion factor (a � hai)/(s(a � hai)) as a function of(a) the proportion of diffuse irradiance f, (b) the normalized net irradiance L/Lmax, and (c) the normalizedvariations of the measured ground relative humidity at IPSL-LMD, (RH � hRHi)/s(RH).

Figure 12. Nighttime ecosystem respiration flux retrievals, NEEnight, as a function of the mean airtemperature. NEEnight is separated according to the season: April–May corresponds to the spring and thegrowing season while July–September to the summer.

D10301 GIBERT ET AL.: CO2 FLUX

11 of 16

D10301

of wind direction, are displayed in Figure 13. Large values ofh close to 0.1 (10 ppb of CO per parts per million of fossilCO2) are obtained when the wind blows from the northeast.Under these conditions, the IPSL-LSCE site lies within theParis pollution plume. Avalue of 10 ppb ppm�1 for the CO tofossil CO2 mixing ratio is close to the one determined for cartraffic emissions by Braud et al. [2004], which lends supportto our ABL method to separate fossil from natural contribu-tions (and fluxes).

7. Uncertainty Analysis

7.1. Error Analysis on ABL Height Estimation

[40] The accuracy of M and NEE is linked to the one ofABL height estimates from lidar measurements and radio-soundings [equations (1)-(3), (5)-(6)]. To study the NBLheight (hNBL) uncertainties, we compared coincident radio-soundings from LMD and Trappes between 18 May and8 June. The temporal error was estimated by the variance ofhNBL at each site during one night. The spatial error wasestimated from the variance of the NBL and ML heightscalculated from distant radiosoundings at the same time.Table 2 summarizes the different error sources on the ABLheight retrieval. The relative error on hNBL is larger com-pared with that on the ML height. In the potential temper-ature and lidar signals, the limit NBL-RL is usually lessmarked than the RL-FA interface which is marked by astrong temperature inversion. During the night, the NBLheight may vary in thickness by displacing the bottom of the

RL. These variations are linked to horizontal wind speedvariations and wave motions [Stull, 1988]. A small increasein hNBL during the night hence explains the large uncertainty(±15%) caused by temporal variability (Table 2).

7.2. Error Analysis on NEEnight Retrieval

[41] To link the linear increase of nighttime CO2 mixingratio and the NEEnight, we need to know the verticalgradient of CO2 into the NBL. In this study, we make thehypothesis that all NBL scalars (CO2, H2O. . .) follow thesame gradient than the vertical potential temperature. Inorder to quantify the possible errors brought by thishypothesis, we chose to compare experimental and modeled[using the potential temperature profile and equation (6)]specific humidity profiles (Figure 14a). The relative erroron the mean specific humidity in the NBL is given inFigure 14b. Except for 4 days among the 21 studied, theuse of the NBL potential temperature profile to predictthe specific humidity one causes an error of less than10%.[42] The relative error on NEEnight also depends on hNBL

and a. To estimate the errors on these parameters, weassumed an exponential decrease of CO2 with height:

C z; tð Þ ¼ C 0; tð Þ � CRL½ exp �z=lð Þ ð7Þ

where l � hNBL/3 and C(0, t) is the CO2 mixing ratio fromground-based in situ measurement.[43] Using equations (5) and (7), we obtain:

NEEnight¼rld

dt

ZhNBL0

C 0; tð Þ�CRLð Þexp�z=lð Þdz

24

35 ð8Þ

Then,

NEEnight ¼ rldC 0; tð Þ

dt1� exp �hNBL=lð Þ½ ð9Þ

The relative error on the respiration flux is:

s NEEnight

� NEEnight

¼ s að Þa

þ 3s hNBLð ÞhNBL

e�3

1� e�3

� �ð10Þ

where a = dC(0, t)/dt is the slope calculated during themorning and the evening periods (Figure 3). The error onthe slope s(a)/a is mainly due to anthropogenic emissionsduring the night. Equation (10) shows that the error on hNBLhas only a marginal influence in the total uncertainty (�2%for 10% on hNBL). This is due to the vertical attenuation ofCO2 with height. At the NBL-RL interface, the CO2 mixing

Table 2. Relative Error on ABL Height Retrievala

ML Height NBL Height

LidarRS: Spatial Error,LMD-Trappes

RS: Temporal error RS: Spatial Error,LMD-TrappesLMD Trappes

d(h)/h 0.05 0.05 0.17 0.12 0.11aLidar backscatter signal and radiosoundings at both facilities, LMD and Trappes, were used to estimate the spatial and

temporal representativity of the ABL height. LMD and Trappes facilities are separated by 10 km.

Figure 13. Estimates of regional fossil CO2/CO emissionratios, h (in parts per million/parts per billion), as a functionof the mean ABL wind direction (in degrees).

D10301 GIBERT ET AL.: CO2 FLUX

12 of 16

D10301

Figure 14. (a) Experimental and modeled (from potential temperature profile) vertical specific humidityprofiles in the NBL using Trappes 7 September-2300 UT radiosounding. (b) Relative error on the meanNBL specific humidity using potential temperature profile along the year 2004.

Figure 15. (a) Statistical relative error on NEEnight calculations from slope estimates during the morningand the evening. (b) Systematic error on NEEnight due to a correction of CO2 ground-based in situmeasurements from anthropogenic emission, assuming a 1 ppm of anthropogenic CO2 from 10 ppb ofCO. (c) Statistical error on b calculation from the linear fit taken during the morning period (see alsoFigure 3). (d) Systematic error on b due to the trapping of anthropogenic emissions of CO2 in the NBLduring the night and assuming a 1 ppm of CO2 from 10 ppb of CO.

D10301 GIBERT ET AL.: CO2 FLUX

13 of 16

D10301

ratios are similar. Therefore NEEnight estimate is only weaklysensitive to variations in hNBL. Then, we obtain:

s NEEnight

� NEEnight

� s að Þa

ð11Þ

[44] Figure 15a shows the relative error on NEEnight forthe morning and the evening periods.[45] The error on NEEnight is on the order of �10%. The

evening period is often contaminated by car traffic emis-sions which add a bias on the slope a (Figure 15b). However,

this error remains below �15%. During the night, theanthropogenic emissions induce an error that is close to thestatistical error in the least squares fit. Thus the value of aproves to be rather insensitive to the anthropogenic emissions(see also Figure 3).

7.3. Uncertainty on the First Point: ddddC(0, 0)[46] The first point of modeled CO2 diurnal cycle C(0, 0)

equals the intercept b of the linear fit to the CO2 datadisplayed (Figure 3). Anthropogenic emissions influencethe value of b. The statistical error on b is small (<1 ppm)

Table 3. Initial and Prescribed Error Values for the Input Parameters h, R, C(0, 0), and C+ and Result Errors on

the Outputs C, P, and M After Monte-Carlo Simulations

Inputs Outputs

s(h)/h s (b)/b s(C(0, 0)), ppm s(C+), ppm s(C), ppm s (NEEday)/NEEday s(M)/M

0.05 – – – <0.3 0.02 0.09– 0.10 – – <1 0.15 0.06– – 2.0 – <2 0.06 0.09– – – 0.5 <0.3 0.08 0.070.05 0.10 2.0 0.5 <2.5 0.16 0.13

Figure 16. Monte-Carlo simulations modeling the 30 July case and assuming that s(C(0, 0)) = 2 ppm,s(b)/b = 0.1, s(h)/h = 0.05, and s(C+) = 0.5 ppm. Initial inputs (a–c) and standard deviation (d–f) areindicated in dark and bold solid lines, respectively. (a) Ground-based CO2 mixing ratio diurnal cycle.(b) NEE, net ecosystem exchange. (c) M, vertical mixing flux. (d–f) Errors on C, NEEday, and M dueto Monte-Carlo simulations.

D10301 GIBERT ET AL.: CO2 FLUX

14 of 16

D10301

(Figure 15c). However, if the NBL traps anthropogenicemissions of the day before, the value of b would be higher.Assuming an average ratio h of 0.1, the effect of anthropo-genic sources on the value of the intercept (parameter db)can be estimated (Figure 15d). The mean value of db is3.5 ppm. In total, we estimate the total error on C(0, 0) to beless than 2 ppm, owing mostly to the systematic error on dbcaused by anthropogenic emissions of the day before.

7.4. Monte-Carlo Analysis of Errors on M and NEEday

Fluxes

[47] Monte-Carlo simulations were performed to analyzethe sensitivity ofM andNEEday to themodel input parameters(Figure 5). The values for b, h, C(0, 0), and C+ are chosenrandomly with a Gaussian distribution of half-width equal toeach parameter’s standard deviation (Table 3). The initial andprescribed CO2 are those of 30 July (see Figure 16a). Themean value of b is prescribed to 7 mmol m�2 s�1 (equaled tothe nighttime NEE).[48] Although the uncertainty onC(0, 0) appears to be large,

it has a small impact on the inference of NEEday [±6% for anerror onC(0, 0) of ppm].Actually, this offset is usually reducedduring the ABL rise in the morning by vertical mixing with airaloft in the RL and FA. Therefore an error on C(0, 0) entails asignificant error on M [±9% for s(C(0, 0)) = 2 ppm]. Theretrieval of M is also sensitive to errors on the CO2 mixingratio in the RL and FA. A random error on C+ estimated to be±0.5 ppm adds an error of 7% to the estimated value of M.Finally, the values of NEEday and NEEnight, i.e., b, are highlycorrelated. For a 10% relative error onR, the relative errors onNEEday and M are of 15 and 6%, respectively (Table 3).[49] Assuming 1 � s Gaussian errors for s(C(0, 0)) = 2 ±

s(b)/b = 0.1, s(h)/h = 0.05, and s(C+) = 0.5 ppm, we made100 Monte-Carlo simulations. The results are given inFigure 16 and Table 3. The total error estimate on M is13% mainly due to uncertainties in ABL height and indC(0, 0). The error on the modeled CO2 diurnal variation isalso due to uncertainties in dC(0, 0). Finally, the relativeerrors on NEEday is found to be on the order of 15%.

8. Conclusion

[50] A method based on boundary layer budgeting andobservations from various in situ and remote sensing instru-ments has been developed to infer regional NEE in presenceof nearby anthropogenic sources. Measurements of CO2

within and above the ABL, in conjunction with a preciseknowledge of the change in ABL height by lidar, enabled toinfer the dependency of NEE on solar radiation. Theretrieved seasonal course of NEE follows the course ofthe growing season, whereas the entrainment flux of CO2 israther linked to the heating of the surface. The diurnal fluxvariations show that the morning decrease of CO2 mixingratio is mainly due to a vertical mixing between the NBLand the residual layer first and then with the free tropo-sphere. Sources of 5 mmol m�2 s�1 during the night andsinks of �20 mmol m�2 s�1 for NEE in the middle of theday in May are in good agreement with nearby eddy-covariance data. The light conversion factor was found toincrease with a larger fraction of diffuse solar radiation.[51] Lidar backscatter proved to be very user changes in

ABL height. One direction for further work is to improve

the capability of the lidar to profile the lower part of theABL near the surface. This will provide some informationabout height and gradients in the NBL that could becompared with vertical distribution of various scalars ofinterest such as potential temperature and specific humidity.[52] This study also opens the way to direct measurement

of CO2mesoscale fluxes using a DIAL system. Indeed, DIALinstrument can simultaneously provide mean CO2 measure-ments in the ABL, aerosol backscatter signal to retrieve thevertical structure of the ABL and velocity information.

[53] Acknowledgments. We thank Pierre Cellier (INRA, Grignon,France) for providing the INRA flux measurements and Jun-Ichi Yano(CNRM, Meteo-France, Toulouse, France) and Kenneth J. Davis(The Pennsylvania State University, USA) for critically reading the man-uscript.

ReferencesAnthoni, P. M., A. Knohl, C. Rebmann, A. Freibauer, M. Mund, W. Ziegler,O. Kolle, and E.-D. Schulze (2004a), Forest and agricultural land-use-dependent CO2 exchange in Thuringia, Germany, Glob. Change Biol.,10, 2005–2019.

Anthoni, P. M., A. Freibauer, O. Kolle, and E.-D. Schulze (2004b), Winterwheat carbon exchange in Thuringia, Germany, Agric. For. Meteorol.,121(1–2), 55–67.

Baldocchi, D., R. Valentini, S. Running, W. Oechel, and R. Dahlman(1996), Strategies for measuring and modelling carbon dioxide and watervapour fluxes over terrestrial ecosystems, Glob. Change Biol., 2, 159–168.

Bakwin, P. S., P. P. Tans, B. B. Stephens, S. C. Wofsy, C. Gerbig, andA. Grainger (2003), Strategies for measurements of atmospheric columnmeans of carbon dioxide from aircraft using discrete sampling, J. Geophys.Res., 108(D16), 4514, doi:10.1029/2002JD003306.

Bakwin, P. S., K. J. Davis, C. Yi, S. C. Wofsy, J. W. Munger, L. Haszpra,and Z. Barcza (2004), Regional carbon dioxide fluxes from mixing ratiodata, Tellus, 56B, 301–311.

Braud, H., P. Bousquet, and M. Ramonet (2004), CO/CO2 Ratio in UrbanAtmosphere: Example of the Agglomeration of Paris, France, no. 42,Institut Pierre Simon Laplace (IPSL), Notes des Activites Instrumentales(N.A.I), Paris.

Crisp, D., et al. (2004), The Orbiting Carbon Observatory (OCO) Mission,Adv. Space Res., 34, 700–703, Manuscript No. A1.2-0007-02.

Davis, K. J., P. S. Bakwin, C. Yi, B. W. Berger, C. Zhaos, R. M. Teclaw,and J. G. Isebrands (2003), The annual cycles of CO2 and H2O exchangeover a northern mixed forest as observed from a very tall tower, Glob.Change Biol., 9, 1278–1293.

de Arellano, J. V.-G., B. Gioli, F. Miglietta, H. J. J. Jonker, H. K. Baltink,R. W. A. Hutjes, and A. A. M. Holtslag (2004), Entrainment process ofcarbon dioxide in the atmospheric boundary layer, J. Geophys. Res., 109,D18110, doi:10.1029/2004JD004725.

Dolman, A. J., R. Ronda, F. Miglietta, and P. Ciais (2005), Regionalmeasurement and modelling of carbon balances, in The Carbon Balanceof Forest Biomes, edited by H. Griffiths and P. G. Jarvis, Taylor andFrancis, Philadelphia, Pa.

Gerbig, C., J. C. Lin, S. C. Wofsy, B. C. Daube, A. E. Andrews,B. B. Stephens, P. S. Bakwin, and C. A. Grainger (2003), Toward cons-training regional-scale fluxes of CO2 with atmospheric observations over acontinent: 2. Analysis of COBRA data using a receptor-oriented frame-work, J. Geophys. Res., 108(D24), 4756, doi:10.1029/2002JD003018.

Gibert, F., P. H. Flamant, D. Bruneau, and C. Loth (2006), Two-micrometerheterodyne differential absorption lidar measurements of the atmosphericCO2 mixing ratio in the boundary layer, Appl. Opt., 45, 4448–4458.

Gloor, M., S.-M. Fans, S. Pacala, J. Sarmiento, and M. Ramonet (1999), Amodel-based evaluation of inversions of atmospheric transport, usingannual mean mixing ratios, as a tool to monitor fluxes of nonreactivetrace substances like CO2 on a continental scale, J. Geophys. Res.,104(12), 14,245–14,260.

Granier, A., K. Pilegaard, and N. O. Jensen (2002), Similar net ecosystemexchange of beech stands located in France and Denmark, Agric. For.Meteorol., 114, 75–82.

Gu, L., D. Baldocchi, S. B. Verma, T. A. Black, T. Vesala, E. M. Falge,and P. R. Dowty (2002), Advantages of diffuse radiation for terrestrial eco-system productivity, J. Geophys. Res., 107(D6), doi:10.1029/2001JD001242.

Hellinker, B. R., J. A. Berry, A. K. Betts, P. S. Bakwin, K. J. Davis,A. S. Denning, J. R. Ehleringer, J. B. Miller, M. P. Butler, andD. M. Ricciuto (2004), Estimates of net CO2 flux by application of

D10301 GIBERT ET AL.: CO2 FLUX

15 of 16

D10301

equilibrium boundary layer concepts to CO2 and water vapor measure-ments from a tall tower, J. Geophys. Res., 109, D20106, doi:10.1029/2004JD004532.

Inoue, G. (2005), The greenhouse gases monitoring in-situ and from space(GOSAT), paper presented at 13th Coherent Laser Radar Conference,Kamakura, Japan.

Katterer, T., M. Reichstein, O. Andren, and A. Lomander (1998), Tempera-ture dependence of organic matter decomposition: A critical review usingliterature data analysed with different models, Biol. Fertil. Soils, 27,258–262.

Levy, P. E., A. Grelle, A. Lindroth, M. Molder, P. G. Jarvis, B. Kruijt, andJ. B. Moncrieff (1999), Regional scale CO2 fluxes over central Sweden bya boundary layer budget method, Agric. For. Meteorol., 98-99, 169–180.

Lloyd, J., and J. Taylor (1994), On the temperature dependence of soilrespiration, Funct. Ecol., 8, 315–323.

Lloyd, J., et al. (2001), Vertical profiles, boundary layer budgets, andregional flux estimates for CO2 and its 13C/12C ratio and for water vaporabove a forest/bog mosaic in central Siberia, Glob. Biogeochem. Cycles,15(2), 267–284.

Menut, L., C. Flamant, J. Pelon, and P. H. Flamant (1999), Urban boundarylayer height determination from lidar measurements over the Paris area,Appl. Opt., 38, 945–954.

Miglietta, F., B. Gioli, R. W. A. Hutjes, and M. Reichstein (2006), Netregional ecosystem CO2 exchange from airborne and ground-based eddy-covariance, land-use maps and weather observations, Glob. Change Biol.,doi:10.1111/j.1365-2486.2006.01219.

Min, Q. (2005), Impact of aerosols and clouds on forest-atmosphere carbonexchange, J. Geophys. Res., 110, D06203, doi:10.1029/2004JD004858.

Moureaux, C., A. Debacq, B. Bodson, B. Heinesch, and M. Aubinet (2006),Annual net ecosystem carbon exchange by a sugar beet crop, Agric. For.Meteorol., 139, 25–39.

Pepin, L., M. Schmidt, M. Ramonet, D. Worthy, and P. Ciais (2002), A NewGas Chromatographic Experiment to Analyze Greenhouse Gases inFlask Samples and in Ambient Air in the Region of Saclay, no. 12, InstitutPierre Simon Laplace (IPSL), Notes des Activites Instrumentales (N.A.I),Paris.

Perrin, D., E. Laitat, M. Yernaux, and M. Aubinet (2004), Modelling of theresponse of forest soil respiration fluxes to the main climatic variables,Biotechnol. Agron. Soc. Environ., 8, 15–25.

Reichstein, M., et al. (2003), Modelling temporal and large-scale spatialvariability of soil respiration from soil water availability, temperature andvegetation productivity indices, Glob. Biogeochem. Cycles, 17(4), 1104.

Stull, R. B. (1988), An Introduction to Boundary Layer Meteorology,Springer, New York.

Turner, D. P., S. V. Ollinger, and J. S. Kimball (2004), Integrating remotesensing and ecosystem process models for landscape to regional scaleanalysis of the carbon cycle, BioScience, 54, 573–584.

Valentini, R., et al. (2000), Respiration as a main determinant of carbonbalance in European forests, Nature, 404, 861–865.

Wofsy, S. C., M. L. Goulden, J. M. Munger, S. M. Fan, and P. S. Bakwin(1993), Net exchange of CO2 in a mid-latitude forest, Science, 260,1314–1317.

Worthy, D. E. J., I. Levin, N. B. A. Trivett, A. J. Kuhlmann, J. F. Hopper,and M. K. Ernst (1998), Seven years of continuous methane observationsat a remote boreal site in Ontario, Canada, J. Geophys. Res., 103(D13),15995–16007.

Yi, C., K. J. Davis, P. S. Bakwin, B. W. Berger, and L. C. Marr (2000),Influence of advection on measurements of the net ecosystem-atmosphereexchange of CO2 from a very tall tower, J. Geophys. Res., 105, 9991–9999.

Yi, C., K. J. Davis, P. S. Bakwin, A. S. Denning, N. Zhang, A. Desai,J. C. Lin, and C. Gerbig (2004), Observed covariance between eco-system carbon exchange and atmospheric boundary layer dynamics ata site in northern Wisconsin, J. Geophys. Res., 109, D08302, doi:10.1029/2003JD004164.

�����������������������P. Ciais, M. Ramonet, M. Schmidt, and I. Xueref, Institut Pierre Simon

Laplace, Laboratoire des Sciences du Climat et de l’Environnement, UMRCEA/CNRS 1572, C.E. de l’Orme des Merisiers, 91191, Gif-sur-Yvette,Cedex, France.J. Cuesta, P. H. Flamant, and F. Gibert, Institut Pierre Simon Laplace,

Laboratoire de Meteorologie Dynamique, Ecole Polytechnique, 91128,Palaiseau, Cedex, France. ([email protected])E. Larmanou, INRA Unite Mixte de Recherche INRA/INAPG

‘‘Environnement et Grandes Cultures’’, 78850, Thiverval-Grignon, France.

D10301 GIBERT ET AL.: CO2 FLUX

16 of 16

D10301